Bug #619
Segfault in @GPythonTestSuite.test_value()@
| Status: | New | Start date: | 12/05/2012 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: |  Knödlseder Jürgen | % Done: | 0% | |
| Category: | - | |||
| Target version: | - | |||
| Duration: |
Description
I was playing with GPythonTestSuite on the command line to see how it works and what it can do, and got this segfault:
$ cat test_GPythonTestSuite.py import gammalib test_suite = gammalib.GPythonTestSuite() test_suite.test_value(0, 0) $ python test_GPythonTestSuite.py Segmentation fault: 11
Is this a bug or am I supposed to not call it with Python integers as input?
In [3]: test_suite.test_value?
Type: instancemethod
String Form:<bound method GPythonTestSuite.test_value of <gammalib.test.GPythonTestSuite; proxy of <Swig Object of type 'GPythonTestSuite *' at 0x10f3a74b0> >>
File: /Users/deil/code/install/lib/python2.7/site-packages/gammalib/test.py
Definition: test_suite.test_value(self, *args)
Docstring:
test_value(GTestSuite self, int const & value, int const & expected, std::string const & name="", std::string const & message="")
test_value(GTestSuite self, int const & value, int const & expected, std::string const & name="")
test_value(GTestSuite self, int const & value, int const & expected)
test_value(GTestSuite self, double const & value, double const & expected, double const & eps=1.0e-10, std::string const & name="",
std::string const & message="")
test_value(GTestSuite self, double const & value, double const & expected, double const & eps=1.0e-10, std::string const & name="")
test_value(GTestSuite self, double const & value, double const & expected, double const & eps=1.0e-10)
test_value(GTestSuite self, double const & value, double const & expected)
Recurrence
No recurrence.
History
#1
Updated by Knödlseder Jürgen about 12 years ago
A segmentation fault is always a bug 
Yet, the class is not intended to be used directly, but it should serve as base class from which the test classes should be derived. Yet I had to make it non abstract.
I have to track done what exactly is happening when the class is used directly.
For the record, here is the crash report:
Exception Type: EXC_BAD_ACCESS (SIGSEGV) Exception Codes: KERN_INVALID_ADDRESS at 0x0000000000000000 Crashed Thread: 0 Dispatch queue: com.apple.main-thread Thread 0 Crashed: Dispatch queue: com.apple.main-thread 0 libstdc++.6.dylib 0x000000010029b8eb std::string::assign(std::string const&) + 11 1 libgamma.0.dylib 0x000000010111c83c GTestSuite::test_value(int const&, int const&, std::string const&, std::string const&) + 1068 2 _test.so 0x0000000101aec822 _wrap_GTestSuite_test_value + 2162 3 org.python.python 0x00000001000b7cb2 PyEval_EvalFrameEx + 22626 4 org.python.python 0x00000001000ba215 PyEval_EvalCodeEx + 2197 5 org.python.python 0x00000001000b81ad PyEval_EvalFrameEx + 23901 6 org.python.python 0x00000001000ba215 PyEval_EvalCodeEx + 2197 7 org.python.python 0x00000001000ba336 PyEval_EvalCode + 54 8 org.python.python 0x00000001000dec2c PyRun_InteractiveOneFlags + 380 9 org.python.python 0x00000001000dee8e PyRun_InteractiveLoopFlags + 78 10 org.python.python 0x00000001000df701 PyRun_AnyFileExFlags + 161 11 org.python.python 0x00000001000eef9d Py_Main + 2909 12 org.python.python 0x0000000100000f14 0x100000000 + 3860
#2
Updated by Deil Christoph about 12 years ago
Jürgen Knödlseder wrote:
A segmentation fault is always a bug
I didn’t know that. So swig always does type checking on inputs?
Apparently it is possible to create gammalib Python classes / methods that do crash.
So an `api-sanity-checker` on the Python side would be nice to catch all the trivial ways to make gammalib crash from Python.
I couldn’t find such a tool, but Jürgen, if you agree that would be useful I’ll make a ticket for myself and find or write something in the future.
#3
Updated by Deil Christoph about 12 years ago
- Assigned To set to Knödlseder Jürgen
#4
Updated by Knödlseder Jürgen about 12 years ago
Christoph Deil wrote:
Jürgen Knödlseder wrote:
A segmentation fault is always a bug
I didn’t know that. So swig always does type checking on inputs?
Apparently it is possible to create gammalib Python classes / methods that do crash.
So an `api-sanity-checker` on the Python side would be nice to catch all the trivial ways to make gammalib crash from Python.
I couldn’t find such a tool, but Jürgen, if you agree that would be useful I’ll make a ticket for myself and find or write something in the future.
This sounds like a good idea. The Python interface is not yet very well tested, a lot of unit tests are still missing. As most of the functionality is tested using the C++ unit tests, I was thinking to focus mostly on the interface for the Python tests, but as you will see, little is done so far.
So a Python sanity checker would certainly help to improve the code quality.